CN104508692B - The device and method adjusted for automatic filter - Google Patents
The device and method adjusted for automatic filter Download PDFInfo
- Publication number
- CN104508692B CN104508692B CN201380039242.5A CN201380039242A CN104508692B CN 104508692 B CN104508692 B CN 104508692B CN 201380039242 A CN201380039242 A CN 201380039242A CN 104508692 B CN104508692 B CN 104508692B
- Authority
- CN
- China
- Prior art keywords
- project
- list
- characteristic value
- factor
- processing unit
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/24—Querying
- G06F16/245—Query processing
- G06F16/2457—Query processing with adaptation to user needs
- G06F16/24578—Query processing with adaptation to user needs using ranking
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
- G06Q30/0251—Targeted advertisements
- G06Q30/0269—Targeted advertisements based on user profile or attribute
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
- H04N21/45—Management operations performed by the client for facilitating the reception of or the interaction with the content or administrating data related to the end-user or to the client device itself, e.g. learning user preferences for recommending movies, resolving scheduling conflicts
- H04N21/466—Learning process for intelligent management, e.g. learning user preferences for recommending movies
- H04N21/4662—Learning process for intelligent management, e.g. learning user preferences for recommending movies characterized by learning algorithms
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
- H04N21/45—Management operations performed by the client for facilitating the reception of or the interaction with the content or administrating data related to the end-user or to the client device itself, e.g. learning user preferences for recommending movies, resolving scheduling conflicts
- H04N21/466—Learning process for intelligent management, e.g. learning user preferences for recommending movies
- H04N21/4668—Learning process for intelligent management, e.g. learning user preferences for recommending movies for recommending content, e.g. movies
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
- H04N21/47—End-user applications
- H04N21/482—End-user interface for program selection
- H04N21/4826—End-user interface for program selection using recommendation lists, e.g. of programs or channels sorted out according to their score
Abstract
The present invention relates to a kind of method and apparatus for adjusting filter parameter, and wherein the device includes: display, physical user interface, memory and the processing unit for being operably connected to display, physical user interface and memory.The memory includes (multiple) classification of the items list, the classification of the items table includes multiple projects in an orderly manner, wherein the sequence of project is determined by its grade, and each project by expression project feature value at least one characteristic value to characterization.The processing unit is configured, to generate the graphical representation of the project in the list in an orderly manner on the display.The processing unit is further configured, to respond to the physical user interface, so that user be allowed to resequence (rearranging) in the graphical representation of the bulleted list and/or abandon project.The processing unit is further configured, to respond physical user interface, after rearranging, according to the graphical representation, modifies the grade of the project in the list.Further configure the processing unit, degree is liked with the determining at least some characteristic values pair for characterizing the project in rearrangement list, with when compared with the characteristic value of the sundry item in the rearrangement list is to the product for liking degree instruction factor, the ratings match of the characteristic value of specific project to the project in the product and rearrangement list for liking degree instruction factor.
Description
Technical field
Recommender system is faced in various application fields for recommended project (product, TV, song etc.), for slowing down
The select permeability of the user of selection is carried out in huge amount option.There are two types of universal methods to construct recommender system.At the first
In method, the project indicated by multiple features and user preferences is according further to these feature representations.This method is commonly known as based on
The recommendation of content.Alternatively, history is listened in the purchase/checking/for analyzing (can be with subsidiary assessment information) large numbers of users,
With the likelihood between identification project or the likelihood between user.Then, new item is recommended to user using these likelihoods
Mesh.The second method is commonly known as collaborative filtering.The new projects suggested in collaborative filtering method are the items liked with user
The similar new projects of mesh, are also possible to the new projects liked with user as given user class.It note that in addition to multiple users'
It buys/checking/and listens to information, which does not need the specific information about project itself.
In general, recommender system is as the filter for being filtered to the possible interested project of user.In order to
This filter is set to be suitble to the expectation and requirement of user, the known method for example based on the sum of content based on collaborative filtering.
It is well known that in both methods of the sum based on content based on collaborative filtering, it is difficult to provide new user good
It is good to recommend.Before recommended device can learn its hobby and can provide good recommend, new user must firstly evaluate big quantifier
Mesh, wherein assessment two aspect (like/do not like) or in terms of with more main points (e.g., including
Five aspects for liking degree do not like very much such as, do not like, is neutral, liking, enjoying a lot) it carries out.
The problem of creation or modification filter parameter prevent them to indicate the hobby of specific users be machine from with
Talk with to understand user preferences at family.Another problem is the lattice for needing to be capable of handling with machine about the information of user preferences
Formula.However, general user cannot understand or revise the format.Can be easily easily absorbed on a cognitive level by the user therefore, it is necessary to one kind and
The user interface for indicating the data of user preferences can effectively be acquired.This is the friendship by cannot directly provide the user of this information
Mutually the basic technology of information can be handled by leading to the problem of machine.
Being learnt by the United States Patent (USP) 7,836,057 that on November 16th, 2010 authorizes requires user to grade project again.?
In United States Patent (USP) 7,836,057, it is proposed that a kind of such as to purchase the method/system for helping user to select product in vehicle.User is bright
Many selection criteria relevant to given product type are really provided, and for every kind of selection criteria, in product selection course
User any degree can be weighted to given standard using sliding block setting.Therefore, system returns to the list of the product of sequence, should
Sort the weight based on selection criteria.Then, if user is dissatisfied to the sequence of product, user can be again to product
List grading.Then, system is pointed out to obtain how the list graded again can adjust the weight of selection criteria.
The method suggested in United States Patent (USP) 7,836,057 has many defects, so that not adapting to more complicated judgement
Journey such as sees any TV or what program is rented in video on demand library.In order to learn the taste of new TV user, and set
Fixed a small amount of selection criteria is compared, and recommender system is more complicated.If studied using Naive Bayes Classification method, for compared with
Multiple characteristic values pair can indicate that the taste of user (refers to the 2007ACM meeting report about recommender system using degree is liked
It concentrates, Pronk, V., W.Verhaegh, the Incorporating User Control into of A.Proidl and M.Tiemann
Recommender Systems Based on Naive Bayesian Classification,RecSys 2007,pp.73–
80,Minne-apolis,MN,USA)。
Term used in the paper, especially with like degree and anti-factor claimed to be suitable for the disclosure.The two
Term has following relationship.It is assumed that r be it is anti-claim factor, then correlation is liked degree λ and is provided by λ=r/ (1+r).On the contrary, for given
Like degree λ, related anti-title factor r is provided by r=λ/(1- λ).0.5 degree of liking represents neutrality, because it is produced as 1
Neutral anti-title factor.The possible range for liking degree is between 0 and 1, and the possible anti-range for claiming factor is in 0 and infinity
Between.
Characteristic value is to can be with some performer or particular genre for occurring in for example given film or specific broadcast
Time or specific broadcast channel are related.Offer does not sound feasible to the user interface of each offer sliding block of these characteristic value centerings
Border.In addition, if including very more characteristic values pair, then user cannot be made to keep scanning the weight of all different characteristic values pair,
And manually setting these is also unpractical selection.In addition, how in order to realize that the given of given list is commented again
Grade does the same difficulty of explanation come the clearly feedback for adjusting weight and user.
The additional defect of United States Patent (USP) 7,836,057 is how to adjust selection according only to the list feedback individually graded again
The weight of standard.The hobby of user is captured to generate television recommendations, single bulleted list too limits, so that cannot learn
Constitute the slight change of user's taste.Hobby that user is captured to generate television recommendations needs a series of to grade again
Step suitably selects the continuous list for requiring user to grade again in those steps.
For learning the taste of TV viewer, another defect of United States Patent (USP) 7,836,057 is to require user to this
All items present in list are graded again.Quantity for carrying out the possibility project of selection by it is very more to answer
With with the television-viewing that can carry out selection from the TV performance of current broadcast, video on demand content, YouTube film etc.
The case where person, is identical, it is desirable that the given list that user grades again be likely to containing user it is ignorant one or more
Mesh.Even if providing the additional information about project, such as type, user is still difficult to grade to it.Therefore, it is proposed that user exists
Again it grades and deletes its ignorant project (for example, replacing using sundry item) before the list from table first.
Another defect of United States Patent (USP) 7,836,057 is can not to explicitly point out user in this way and not like in list
Which project.Possible its likes all items, does not like all items, or only likes a m project, wherein 0 < m < n.Cause
This, the another aspect of the preferred embodiment of the present invention is that user can be in lists before first project or at last
After a project between any continuous item pair tab-delimited symbol, to provide between the project liked and the project not liked
Boundary.
Summary of the invention
The object of the present invention is to provide a kind of filterings for allowing the specific needs according to user and/or it is expected filtering items
The user-friendly creation or modification of device.
According to the first aspect of the invention, which is realized by the device for adjusting filter parameter, wherein the device
Including or be connectable to: display, physical user interface and memory.The device includes processing unit, the processing unit
It is operably connected to the display, physical user interface and memory.The memory includes: (multiple) classification of the items list,
The classification of the items list includes multiple projects in an orderly manner, and wherein the sequence of project is determined by its grade.Configure the processing list
Member, to generate the graphical representation of the project in the list in an orderly manner on the display.The processing unit is further configured,
To be responded to the physical user interface, so that user be allowed to resequence (again in the graphical representation of the bulleted list
Arrangement) and/or abandon project.The processing unit is further configured, to respond physical user interface, after rearranging,
According to the graphical representation, the grade of the project in the list is modified.The processing unit is further configured, according to rearrangement
Related evaluation history is modified in list, and the related evaluation history by adjusting generates the filter parameter of one group of adjustment.
The filter parameter of group modification can for example define the user's overview of modification to recommended device.
Initial project tabulation can be classified by some default level that can be created at random.Alternatively, initial point
Class list can according to need supposition user's overview update or that needs are personalized or the classification of expired user's overview.Herein
It is recommended that the benefit of method be that user sees and understand shown project, without know which feature/value to or its
His machine can handle that information is related with shown project, while processing unit can immediately treat the information with the item association
With the information to interact the sequential encoding of the rearrangement or the list reclassified that obtain by user.
Preferably, the processing unit is further configured, to respond to physical user interface, to allow user will be every
A project be regarded as belonging to for example to like or do not like at least two groups in one.Therefore, user will can be shown
Item dividing is two groups, and can generate (absolute) evaluation history.Then, by several classification methods (collaborative filtering, simplicity
Bayes's classification, support vector machines) any one of method be used as project filter (recommended device), this can be further processed
(absolute) evaluation history.These project filters are the classifiers for typically constructing user's overview or model, which makees
For filter parameter work.Equally, related evaluation history can be divided into such as 5 groups, so that each group indicates from 1 point
To the assessment of 5 points of aspect (with desired sequence).This can also and even more suitable for in collaborative filtering environment.
Guarantee that allowing the user interface for dividing project group to mean that from ease of user interaction retrieves more more information,
Wherein in the apparatus, the information retrieved in this way is mutually compatible with the internal representation of user preferences.
According to preferred embodiment, the processing unit is further configured, related evaluation history is divided into two groups: head N
A will be the project liked, and sundry item is the project not liked.That is done as follows further describes, it is preferable that this can be by
The preferred embodiment of processing unit is realized, the processing unit is configured, with allow in the list rearranged first item it
It is preceding or between any pair of continuous item after final race tab-delimited symbol, wherein separator provides the project liked
With the boundary between the project that does not like.The separator defines decision threshold.The preferred embodiment of processing unit can include
Or it is connectable to the interface unit for allowing display and mobile separator.Then, the processing unit is configured, by separator
Position be used as input value for generating the list of filter parameter.Because separator can be arranged in front of first item
Perhaps after final race thus user all display projects in the list can be respectively labeled as " liking " or
" not liking ".
In a preferred embodiment, each project by expression project characteristic value at least one characteristic value to characterization.Cause
This, it is preferable that the processing unit is configured, further to determine the feature of the project in the list for characterizing rearrangement at least
Some characteristic values pair like degree instruction factor, to work as the characteristic value with the sundry item in the list of the rearrangement to happiness
When the product of joyous degree instruction factor is compared, column of the characteristic value of specific project to the product and rearrangement of liking degree instruction factor
The ratings match of project in table.
The grade of project in the list of rearrangement is related to degree is liked, and therefore, and with lower grade
Project is compared, and the degree of liking of the project in the project with higher level indicates that factor may be liked more by specific user
Vigorously.In a preferred embodiment, which determines that the degree of liking of characteristic value pair indicates factor, so that each project (or extremely
Few some projects) characteristic value the sequence that the grade of the project provides is corresponded to the sequence for the product for liking degree instruction factor.
The processing unit can be further configured, to respond user's interaction by physical user interface, is being rearranged
Later, according to graphical representation, the grade of the project in the list is modified.The attribute has for describing the project in rearrangement
What three states have occurred, it may be assumed that (project) is deleted, (project) moves up, (project) moves down.
User quickly provides its hobby in a user-friendly manner for apparatus according to the invention permission, and therefore generates use
In the filter of effective filtering items.
Therefore, it is resequenced multiple bulleted lists properly selected the present invention provides a kind of by requiring user, and
It is capable of the new method of the taste of the new user of Fast Learning.
Preferably, liking degree instruction factor is the anti-title factor of particular characteristic value pair, and characterizes specific project
The product of the anti-title factor of characteristic value pair is the anti-title factor of the project.Due to project by one or more characteristic value to feature
Change, and like that degree is related with project, and the characteristic value pair for characterizing specific project likes degree and the happiness of the project
Joyous degree cross-correlation.Like degree if determined using Naive Bayes Classification, characteristic value to have it is anti-claim factor r, and
The anti-title factor r (x) of project x is by characterizing the characteristic value of the project to the product Π of the anti-title factor of F (x)i∈F(x)riIt gives
Out.In this embodiment, processing unit is preferably used for solving one group of linear inequality Σi∈F(x_j)ρi>Σi∈F(x_(j+1))ρi,
Middle ρiIt is the anti-title factor r of project iiLogarithm log (ri), and wherein x is indicated using symbol x_jj。
In further preferred embodiments, memory includes multiple ordered items lists, and wherein processing unit is used for
Continuously generate the graphical representation of the project in each list in an orderly manner over the display.In such an embodiment, further
The processing unit is configured, to respond physical user interface, so that user be allowed to arrange again in the graphical representation of the bulleted list
Project is arranged and/or abandoned, and responds physical user interface, after rearranging, according to the graphical representation, modifies the column
The grade and attribute of project in table.
In general, multiple projects preferably usually have the characteristic value pair of at least one subset, and wherein to each
Project and each characteristic value like degree to distribution, so that project likes degree by liking degree instruction factor (for example, anti-claim
Factor) definition, it is described like degree instruction factor be characterizes the project characteristic value pair like degree indicate factor
Product.In this respect, it is further preferred that configuring the processing unit, to respond the input that user is keyed in by physical user interface,
After having rearranged the ordered list of project, degree is liked by the rating calculation characteristic value pair of project.
Memory preferably includes multiple classification of the items lists, likes degree with determine more features value pair, without
Want user in face of too long of bulleted list.If showing second item list, preferably, the filter tune with front to user
As determination in section processing, according to the sequence (grade) of characteristic value pair liking degree and identifying project.
Preferably, database is configured by memory.
Preferably, the device is configured, with the seed item selected using such as user by user interface, and configuring should
Processing unit, to generate item according to the degree of likeness between the seed item and the sundry item being stored in the memory
Mesh tabulation.
According to another aspect of the invention, a kind of method for adjusting filter parameter is provided.This method includes step
It is rapid:
Classification of the items list including multiple projects is provided in an orderly manner, wherein the sequence of project is determined by its grade,
And each project by expression project characteristic value at least one characteristic value to characterization,
Graphical representation is generated to the project in the list in an orderly manner over the display,
Physical user interface is responded, to allow user to resequence (rearranging) in the graphical representation of bulleted list
And/or project is abandoned,
Response physical user interface, according to graphical representation, modifies the grade of the project in the list after rearranging,
According to rearrangement list, related evaluation history is modified, and
By the related evaluation history modified, a filter parameters of modification are generated.
According to preferred embodiment or alternate embodiment, this method further includes determining the item characterized in rearrangement list
At least some characteristic values pair of purpose feature like degree instruction factor, to work as and the sundry item in the rearrangement list
Characteristic value when being compared to the product for liking degree instruction factor, the characteristic value of specific project to like the product of degree instruction factor with
The step of ratings match of project in rearrangement list.
According to preferred embodiment, this method is further comprised the steps of:
Using seed item, and
According to the degree of likeness between the seed item and the sundry item being stored in the memory, project point is generated
Class list.
It is further preferred that if this method further comprises the steps of:
According to the degree of liking of the characteristic value pair of determining characterization project, new projects' tabulation is generated.
According to a further preferred embodiment, this method further comprises the steps of:
Graphical representation is generated to the project in new projects' tabulation in an orderly manner over the display,
Physical user interface is responded, to allow user to resequence (again in the graphical representation of new projects' tabulation
Arrangement) and/or abandon project,
Response physical user interface, according to graphical representation, modifies the project in new tabulation after rearranging
Grade, and
Determine at least some characteristic values pair for characterizing the project in rearrangement list likes degree, when heavy with this
When the characteristic value of sundry item in new sort list compares the product for liking degree instruction factor, the characteristic value pair of specific project
Like the product of degree instruction factor and the ratings match of the project in rearrangement list.
Detailed description of the invention
It is discussed in greater detail according to below with reference to following attached drawing to what the present invention was done, above-mentioned and its other party of the invention
Face, feature and advantage become apparent, in which:
Fig. 1 is the graphical representation of the device of the automatic adjustment for filter parameter；
Fig. 2 is the graphical representation of the device in Fig. 1 to project rearrangement；And
Fig. 3 is the graphical representation of the alternative of the automatic adjustment for filter parameter.
Specific embodiment
Device 10 shown in Fig. 1 for the automatic adjustment of filter parameter be connected to display 12 and such as mouse,
The physical user interface 14 of track packet etc..The device includes or is connected to the memory 16 of the tabulation including project x, should
Tabulation includes multiple projects in an orderly manner, and wherein the sequence of project is determined by its grade.Each project is by expression item
At least one characteristic value of purpose characteristic value characterizes i.In general, project x is by multiple characteristic values to (i1、i2、i3...) special
Signization.Degree λ is liked in addition, specified to each project x.The grade of project x liked degree and define the project.16 energy of memory
Enough it is the integration section of device 10, also can is the database that device 10 is connected to.
Device 10 further includes user interface section 18, processing unit 20 and display interface unit 22.
User interface section 18 is configured, to receive signal from physical user interface 14 and corresponding signal is forwarded to processing
Unit 20.
Processing unit 20 is connected to memory 16, and therefore, is able to access that one or more classification of the items list, such as
It is upper described.Processing unit 20 also makes display interface unit 22 generate the graphical representation for leading to classification of the items list on the display 12
Signal.Processing unit 20 is configured, to work as filter adjusts unit, as described below.
Fig. 2 is the expression of the device 10 adjusted for automatic filter, and wherein user arranges again on the display 12
Sequence rearranges project.The grade for rearranging change project of project on the display 12.Because the grade of project with
It likes degree correlation, so different grade most probables causes the difference of project to like degree.Degree is liked due to project
Characteristic value pair depending on characterizing project likes degree, so the variation of the grade of project eventually leads to the happiness of characteristic value pair
The revaluation of joyous degree, the further detailed description done as follows.
In fig. 3 it is shown that a kind of alternative arrangement, wherein memory 16 is not intended to the device of automatic filter adjusting
10 ' integration section, but a part of remote data base.Therefore it provides processing unit 20 is allowed to access in remote data base
Memory 18 data-interface 24.
Now, the operation of device 10 is described.
Basic idea of the invention be provided for new user a kind of television recommender systems or similar recommender system, by
A series of continuous item lists are provided in interactive session, it is specified that its hobby-and therefore reconciling items filter it is easy simultaneously
And convenient method, wherein requiring that user executes following step for each continuous item list.
1. user deletes its project for being not enough to grade to it to the understanding of project, it can use another project replacement and delete
Project；
It grades again 2. user likes reduced sequence with user to the bulleted list of acquisition；
3. in addition, user can be in lists before first project or after the last one project in any company
Tab-delimited symbol between continuous project pair, to provide the boundary between the project liked and the project not liked.
Device utilize include again information in the sequence of grading list adjust appropriate characteristic value pair like degree,
It is middle to combine the information from list of currently grading again with the information from list of previously grading again.Then, it utilizes
The adjustment parameter of the filter of recommender system determines next appropriate project list, wherein typically, continuous list and user's
The matching of taste is become better and better.
For the list of grading again, user is only by the n tables to the end of given position drag and drop in project from the beginning 10 tables
In new demand position.In this way, user, which merely has to the project in correct n, provides opposite user preferences.In psychological study
It is well known that user is easier compared with it once to a project evaluation must provide absolute user preferences to one group of project
Opposite grading is provided to one group of given project.
The list of initial head n inconsistent from personal taste well select however with user of grading again is opened
Begin, deletes and be performed many times with ranking process again.First graded again list according to this, recommended device provide it is next
Attempt n, head that indicate the taste of user slightly goodly lists.Again second n a list of grading provides pass for recommended device
In the additional information of user's taste, in this way in next iteration, the list of n, further good head is obtained.It can be repeatedly this
Again it grades iteration, until user is full to next head n lists of acquisition or to a series of next head n lists
Meaning.For lasting improvement recommended device, can choose continuous head n lists so that single project in these continuous lists most
Often have primary.In addition, continuous list is not also not that should contain item closely similar mutually for example for its associated eigenvalue pair
Mesh.
In this way, user repeats to obtain the feedback for how well learning its taste about recommended device.It is believed that with one time one
Aly only to the single project evaluation, the taste without further how well to learn user to recommended device, which is done, to be fed back, and the process is more
It is helpful.Because more helpful (because repeating to feed back) and being easier to grade again, and replace individually assessing project, so
Always excitation user persistently provides feedback.Therefore, it is recommended that device can provide significant recommendation in earlier stage.
In third step, user can wish that obtaining the project recommended thinks with it to not interested enough the item of recommendation at it
Separator is set between mesh.In this way, user can not only provide opposite grading to one group of given project, and can be in absolute project
In provide its hobby.Response in ordered list between two continuous items i and i+1 tab-delimited symbol, processing unit 20 will
Decision threshold t is set as (λ _ i+ λ _ (i+1))/2.If before separator is located in first item, processing unit 20 will
Decision threshold t is set as (λ _ 1 1+)/2.If separator is located in final race, that is, after project n, then processing unit 20
Decision threshold t is set as λ _ n/2.Classifier can utilize the project that decision threshold difference is liked and the project not liked.
For example, its positive prior probability can be set as such as 1-t by the Naive Bayes Classifier realized by processing unit.Therefore,
Separator correctly to define prior probability present in Naive Bayes Classifier, make classifier also distinguish the project liked with
The project not liked.
In order to adjust " user's overview " (that is, filter filtering items) according to n, the head graded again lists, it is proposed that
The following examples.In this example it is assumed that recommender system uses Nae Bayesianmethod, in the Nae Bayesianmethod
In, directly adjust characteristic value pair related with the project occurred in this n lists likes degree.
Adjust characteristic value pair likes degree
For simplicity, it is assumed that the sequence of n given project assigns each project in n project based on recommended device
That gives likes degree.Assuming that the group all may characteristic value provided to by F={ 1,2 ..., N }.Now, project x can be by son
CollectionIt characterizes.
For each characteristic value to i ∈ F, real number value ri∈ [0, ∞] provides the anti-title factor of this feature value pair, so that item
The skew factor r (x) of mesh x is by ∏i∈F(x)riIt provides.The degree of liking of related more intuition is provided by λ (x) and by r (x)/(1+r
(x)) it provides.
Now, it provides and utilizes x1、x2、……、xnThe grading list again of n project of expression, problem are whether we can
One group of assemblage characteristic value of enough selections and characteristic value pair is to F (x1)∪F(x2)∪……∪F(xn) in characteristic value i is liked
Degree, so that λ (x1)>λ(x2)>...>λ(xn) the sequence for liking degree and project rearrangement (or rearranging) arrange
The sequence of the tier definition of table matches.It is incremented by due to liking degree and the anti-transitions monotonic claimed between factor, so we can be with
In other words this is expressed as whether we can will be selected to be F (x1)∪F(x2)∪...∪F(xn) characteristic value pair anti-title
Factor, so that r (x1)>r(x2)>...>r(xn)。
Then, it is monotonically increasing function using the algorithm, can is as follows one group of linear inequality by the problem reduction.Such as
Fruit we utilize ρiIndicate log (ri), then it can be by inequality r (xj)>r(xj+1) it is rewritten as Σi∈F(x_j)ρi>Σi∈F(x_(j+1))
ρi, wherein indicating x using symbol x_jj。
It can be one group of limited linear equality by the problem representation, if it is present for example utilizing simplex method energy
Enough determine its solution.If there is no the solution, then it is able to carry out search, there is for example minimum dissatisfied solution restricted to find.
The process can be repeated to the several lists suitably selected, so that the major part in characteristic value space is capped, and
And create the more comprehensively overview of user.There are several ways for example to generate the new of the project of height assessment using the overview so far established
List, to create the grading list of randomly selected new projects.Alternatively, according to such as type, it can choose one group of more phase
As program.Another modification is one or more distinguishing characteristics amplification to the overview so far established, for example, to according to height
Like the especially significant feature amplification of degree.
(Pronk, V., J.Korst, M.Barbieri, and A.Proidl.Personal are referred to according to personal channel
television channels:simply zap-ping through your PVR content,in the
Proceedings of the 1st International Workshop on Recommendation-based
Industrial Applications,in conjunction with the 3rd ACM Conference on
Recommender Systems, RecSys 2009, New York City, NY.), user will usually be generated it is many these
Personal channel.It creates personal channel and selects so-called seed programs (seed program) to start typically via user.Now, root
According to the mode of present example, the process that study corresponds to the taste of newly generated personal channel can be implemented, wherein utilizing institute
Choose seeds the building continuous item list of sub- program guide so that by original item be attached to it is similar with given seed programs (very or
Person is slightly) these lists.
The present invention can be applied to use any environment of recommended device, for example, books, song, taxi video etc.
Environment.In addition, it can be used to personal channel environment.Independent recommended device can be made related to each channel, and new user
Problem, the new channel problem being more suitably known as are likely encountered repeatedly.
Claims (13)
1. a kind of for adjusting the device of filter parameter, described device includes or is connectable to:
Display, physical user interface and memory,
Described device further includes processing unit, and the processing unit operationally or can be operably connected to the display
Device, physical user interface and memory,
Wherein the memory includes:
Classification of the items list, the classification of the items list include multiple projects in an orderly manner, and wherein the sequence of project is by its etc.
Grade is determining,
Wherein each project is characterized in that multiple characteristic values pair, and
Wherein, the processing unit is configured to generate the figure of the project in the list in an orderly manner on the display
It indicates,
Wherein, the processing unit is further configured to respond the physical user interface, to allow user in institute
It states and resequences and/or abandon project in the graphical representation of classification of the items list, so that rearrangement list is generated,
Wherein, the processing unit is further configured to response physical user interface, after rearranging, according to figure table
Show, modifies the grade of the project in the list, and
Wherein, the processing unit is further configured to modify related evaluation history, and by modifying according to rearrangement list
Related evaluation history generate one group of adjustment filter parameter,
Wherein, the processing unit is further configured to respond physical user interface, to allow user by each item
Mesh qualification turns to related one liked in degree group for belonging to predetermined quantity, described related to like degree group and include at least to like
First group of project and do not like second group of project,
The processing unit is configured to generate absolute assessment history,
Wherein, characteristic value to indicate project feature value, and wherein the processing unit be further configured to determine feature
That changes at least some characteristic values pair of the project in rearrangement list likes degree instruction factor, so that when arranging again with described
When the characteristic value of sundry item in sequence table compares the product for liking degree instruction factor, the characteristic value of specific project is to liking
Degree indicates the product of factor and the ratings match of the project in rearrangement list, and
Wherein, described to like degree instruction factor and be the anti-title factor of particular characteristic value pair, and wherein characterize specific project
Characteristic value pair anti-title factor product be the project anti-title factor, and
Wherein, the multiple project usually has the subset of at least characteristic value pair, and wherein to each project and each feature
Value likes degree instruction factor to distribution, so that project likes degree instruction factor by the characteristic value pair of the characterization project
Like degree instruction factor product definition.
2. the apparatus according to claim 1, wherein the processing unit is suitable for solving one group of linear inequality ∑i∈F(x_j)
ρi>∑i∈F(x_(j+1))ρi, wherein ρiIt is the anti-title factor r of project iiLogarithm log (ri), and wherein symbol x_j is used to indicate
xj。
3. device according to claim 1 or 2, wherein the memory includes multiple ordered items lists, and wherein
The processing unit is suitable for continuously generating the graphical representation of the project in each list in an orderly manner over the display,
Wherein, the processing unit is further configured to response physical user interface, to allow user in the bulleted list
Project is rearranged and/or abandoned in graphical representation, and responds physical user interface, after rearranging, according to institute
Graphical representation is stated, the grade of the project in the list is modified.
4. the apparatus according to claim 1, wherein the processing unit is configured to response and passes through the defeated of physical user interface
Enter, after the rearranging of ordered list of project, degree instruction factor is liked by the rating calculation characteristic value pair of project.
5. device according to claim 1 or 2, wherein described device is configured to using seed item, and the processing
Unit is configured to be generated according to the degree of likeness between the seed item and storage sundry item in the memory
Classification of the items list.
6. device according to claim 1 or 2, wherein the memory is configured to database.
7. device according to claim 1 or 2, wherein described device is configured to allow in tabulation in first item
Separator is keyed between any pair of continuous item before mesh or after final race, and wherein the processing unit is matched
It is set to response and separator is located between any pair of continuous item before first item or after final race, setting
Decision threshold t.
8. a kind of method for adjusting filter parameter, the method includes the steps:
Classification of the items list including multiple projects is provided in an orderly manner, wherein the sequence of project is determined by its grade, and
Wherein each project is characterized in that multiple characteristic values pair,
Generate the graphical representation of the project in the list in an orderly manner over the display,
Physical user interface is responded, to allow user to resequence and/or abandon in the graphical representation of classification of the items list
Project,
Response physical user interface, according to graphical representation, modifies the grade of the project in the list after rearranging,
According to rearrangement list, related evaluation history is modified, and
By the related evaluation history modified, a filter parameters of adjustment are generated,
The method further includes:
Physical user interface is responded, to allow user that each project qualification is turned to the related happiness for belonging to predetermined quantity
It is one in joyous degree group, described related to like degree group and include at least first group of the project liked and the project not liked
Second group,
Absolute assessment history is generated,
Wherein, characteristic value to indicate project feature value, and wherein the method further includes determine characterize again
At least some characteristic values pair of project in sorted lists like degree instruction factor, so that working as and the rearrangement list
In sundry item characteristic value to like degree instruction factor product compare when, the characteristic value of specific project refers to degree is liked
Show the product of factor and the ratings match of the project in rearrangement list, and
Wherein, described to like degree instruction factor and be the anti-title factor of particular characteristic value pair, and wherein characterize specific project
Characteristic value pair anti-title factor product be the project anti-title factor, and
Wherein, the multiple project usually has the subset of at least characteristic value pair, and wherein to each project and each feature
Value likes degree instruction factor to distribution, so that project likes degree instruction factor by the characteristic value pair of the characterization project
Like degree instruction factor product definition.
9. according to the method described in claim 8, wherein, the method also includes determining the item characterized in rearrangement list
At least some characteristic values pair of purpose like degree instruction factor, so that working as and the sundry item in the rearrangement list
When characteristic value compares the product for liking degree instruction factor, the characteristic value of specific project is to the product for liking degree instruction factor and again
The step of ratings match of project in new sort list.
10. according to method described in claim 8 or 9, wherein the method also includes steps:
Using seed item, and
According to the seed item and the degree of likeness being stored between the sundry item in memory, classification of the items column are generated
Table.
11. according to the method described in claim 8, wherein, the method also includes steps:
According to the filter parameter of modification, new projects' tabulation is generated.
12. according to the method described in claim 9, wherein, the method also includes steps:
Degree instruction factor is liked according to the determination of the characteristic value pair of the project of characterization, generates new projects' tabulation.
13. method described in 1 or 12 according to claim 1, wherein the method also includes steps:
Generate the graphical representation of the project in new projects' tabulation in an orderly manner over the display,
Physical user interface is responded, to allow user to resequence and/or put in the graphical representation of new projects' tabulation
Abandoning project,
Respond physical user interface, after rearranging, according to graphical representation, modify the project in new tabulation etc.
Grade, and
According to rearrangement list, related evaluation history, and the related evaluation history by modifying are modified, generates one group of adjustment
Filter parameter, and/or
Determine at least some characteristic values pair for characterizing the project in rearrangement list likes degree instruction factor, so that working as
When compared with the characteristic value of the sundry item in the rearrangement list is to the product for liking degree instruction factor, specific project
Ratings match of the characteristic value to the project in the product and rearrangement list for liking degree instruction factor.
Applications Claiming Priority (3)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| EP12175382.6 | 2012-07-06 | ||
| EP12175382.6A EP2682910A1 (en) | 2012-07-06 | 2012-07-06 | Device and method for automatic filter adjustment |
| PCT/EP2013/064314 WO2014006209A1 (en) | 2012-07-06 | 2013-07-05 | Device and method for automatic filter adjustment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104508692A CN104508692A (en) | 2015-04-08 |
| CN104508692B true CN104508692B (en) | 2019-10-01 |
Family
ID=48746559
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201380039242.5A Active CN104508692B (en) | 2012-07-06 | 2013-07-05 | The device and method adjusted for automatic filter |
Country Status (6)
| Country | Link |
|---|---|
| US (1) | US9959316B2 (en) |
| EP (2) | EP2682910A1 (en) |
| CN (1) | CN104508692B (en) |
| IN (1) | IN2015DN00832A (en) |
| RU (1) | RU2633096C2 (en) |
| WO (1) | WO2014006209A1 (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140317105A1 (en) * | 2013-04-23 | 2014-10-23 | Google Inc. | Live recommendation generation |
| US10296612B2 (en) | 2015-09-29 | 2019-05-21 | At&T Mobility Ii Llc | Sorting system |
| US10416959B2 (en) | 2015-10-27 | 2019-09-17 | At&T Mobility Ii Llc | Analog sorter |
| US10261832B2 (en) | 2015-12-02 | 2019-04-16 | At&T Mobility Ii Llc | Sorting apparatus |
| US10496370B2 (en) | 2015-12-02 | 2019-12-03 | At&T Intellectual Property I, L.P. | Adaptive alphanumeric sorting apparatus |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5616876A (en) * | 1995-04-19 | 1997-04-01 | Microsoft Corporation | System and methods for selecting music on the basis of subjective content |
| US7836057B1 (en) * | 2001-09-24 | 2010-11-16 | Auguri Corporation | Weighted preference inference system and method |
| CN102265272A (en) * | 2008-12-23 | 2011-11-30 | 阿克塞尔斯普林格数字电视指导有限责任公司 | Biased recommender system |
Family Cites Families (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR0129964B1 (en) * | 1994-07-26 | 1998-04-18 | 김광호 | Musical instrument selectable karaoke |
| US5930784A (en) * | 1997-08-21 | 1999-07-27 | Sandia Corporation | Method of locating related items in a geometric space for data mining |
| US6493702B1 (en) * | 1999-05-05 | 2002-12-10 | Xerox Corporation | System and method for searching and recommending documents in a collection using share bookmarks |
| JP4065381B2 (en) * | 1999-11-10 | 2008-03-26 | ヤフー！ インコーポレイテッド | Internet radio and broadcast method |
| US7937725B1 (en) * | 2000-07-27 | 2011-05-03 | Koninklijke Philips Electronics N.V. | Three-way media recommendation method and system |
| US7757250B1 (en) * | 2001-04-04 | 2010-07-13 | Microsoft Corporation | Time-centric training, inference and user interface for personalized media program guides |
| DE10154656A1 (en) * | 2001-05-10 | 2002-11-21 | Ibm | Computer based method for suggesting articles to individual users grouped with other similar users for marketing and sales persons with user groups determined using dynamically calculated similarity factors |
| US9092523B2 (en) * | 2005-02-28 | 2015-07-28 | Search Engine Technologies, Llc | Methods of and systems for searching by incorporating user-entered information |
| US8412698B1 (en) | 2005-04-07 | 2013-04-02 | Yahoo! Inc. | Customizable filters for personalized search |
| US8468244B2 (en) * | 2007-01-05 | 2013-06-18 | Digital Doors, Inc. | Digital information infrastructure and method for security designated data and with granular data stores |
| JP4941161B2 (en) | 2007-08-02 | 2012-05-30 | ブラザー工業株式会社 | Motion detection device |
| US20090163183A1 (en) * | 2007-10-04 | 2009-06-25 | O'donoghue Hugh | Recommendation generation systems, apparatus and methods |
| US20090164929A1 (en) * | 2007-12-20 | 2009-06-25 | Microsoft Corporation | Customizing Search Results |
| CN101661477A (en) | 2008-08-26 | 2010-03-03 | 华为技术有限公司 | Search method and system |
| US8160996B2 (en) * | 2009-02-02 | 2012-04-17 | The Hong Kong Polytechnic University | Sequence online analytical processing system |
-
2012
- 2012-07-06 EP EP12175382.6A patent/EP2682910A1/en not_active Withdrawn
-
2013
- 2013-07-05 WO PCT/EP2013/064314 patent/WO2014006209A1/en active Application Filing
- 2013-07-05 IN IN832DEN2015 patent/IN2015DN00832A/en unknown
- 2013-07-05 EP EP13734119.4A patent/EP2870578A1/en not_active Ceased
- 2013-07-05 CN CN201380039242.5A patent/CN104508692B/en active Active
- 2013-07-05 US US14/413,069 patent/US9959316B2/en active Active
- 2013-07-05 RU RU2015103735A patent/RU2633096C2/en not_active IP Right Cessation
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5616876A (en) * | 1995-04-19 | 1997-04-01 | Microsoft Corporation | System and methods for selecting music on the basis of subjective content |
| US7836057B1 (en) * | 2001-09-24 | 2010-11-16 | Auguri Corporation | Weighted preference inference system and method |
| CN102265272A (en) * | 2008-12-23 | 2011-11-30 | 阿克塞尔斯普林格数字电视指导有限责任公司 | Biased recommender system |
Also Published As
| Publication number | Publication date |
|---|---|
| US20150169585A1 (en) | 2015-06-18 |
| US9959316B2 (en) | 2018-05-01 |
| IN2015DN00832A (en) | 2015-06-12 |
| EP2870578A1 (en) | 2015-05-13 |
| EP2682910A1 (en) | 2014-01-08 |
| WO2014006209A1 (en) | 2014-01-09 |
| CN104508692A (en) | 2015-04-08 |
| RU2633096C2 (en) | 2017-10-11 |
| RU2015103735A (en) | 2016-08-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104508692B (en) | The device and method adjusted for automatic filter | |
| US20120278342A1 (en) | Intelligent default weighting process for criteria utilized to score media content items | |
| Quintarelli et al. | Recommending new items to ephemeral groups using contextual user influence | |
| CN100492351C (en) | Device for constructing search inquiry and method and system for searching database | |
| US9529801B2 (en) | Multimedia analysis and recommendation generation | |
| CN105893443A (en) | Video recommendation method and apparatus, and server | |
| JP2005505070A (en) | Personal recommendation device database using other people's profiles | |
| Sun et al. | A novel approach for collaborative filtering to alleviate the new item cold-start problem | |
| KR20040075036A (en) | Hierarchical decision fusion of recommender scores | |
| JP2005504399A (en) | Change personal profile by recommender using other person's profile | |
| CN1692352A (en) | User interface providing automatic organization and filtering of search criteria | |
| KR20050120786A (en) | Method and apparatus for grouping content items | |
| CN112235614B (en) | Personalized recommendation method and system for ranking list data of television programs | |
| WO2012134854A1 (en) | Multimedia analysis and cluster formation | |
| KR20170079429A (en) | A clustering based collaborative filtering method with a consideration of users' features and movie recommendation system using thereof | |
| US11036743B2 (en) | Methods, systems, and media for presenting content organized by category | |
| CN114513687A (en) | Server and media asset recommendation method | |
| US20150324446A1 (en) | System and method for searching a labeled predominantly non-textual item | |
| CN103369373B (en) | Method and apparatus for updating content channel | |
| US20180367856A1 (en) | Scalable calculation of the similarity content graph | |
| US20130246441A1 (en) | Method for Evaluating Short to Medium Length Messages | |
| CN103369375A (en) | A method and an apparatus for content channels | |
| CN103369376A (en) | Method and apparatus for content channels using references | |
| CN107025277B (en) | A kind of Quantitative marking method of user concealed feedback | |
| Padmanabhan et al. | A rule based approach to group recommender systems |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20201202 Address after: hamburg Patentee after: Fink TV guidance Co., Ltd Address before: Berlin Patentee before: Fink Digital Television Director Co.,Ltd. |